'dpdkvdev' ports are for use with virtual DPDK Poll Mode Drivers eg. null, pcap. To add a DPDK vdev, a valid 'dpdk-devargs' must be set. The format expected is 'eth_<driver_name><x>' where 'x' is a number between 0 and RTE_MAX_ETHPORTS -1.
For example to add a 'dpdkvdev' port that uses the 'null' DPDK PMD driver: ovs-vsctl set Interface null0 options:dpdk-devargs=eth_null0 All virtual DPDK PMDs haven't been tested so this port type can be considered experimental. Signed-off-by: Ciara Loftus <ciara.lof...@intel.com> --- INSTALL.DPDK-ADVANCED.md | 23 ++++++++++++++++++ NEWS | 1 + lib/netdev-dpdk.c | 62 ++++++++++++++++++++++++++++++++++++++++++++++-- vswitchd/vswitch.xml | 9 +++++-- 4 files changed, 91 insertions(+), 4 deletions(-) diff --git a/INSTALL.DPDK-ADVANCED.md b/INSTALL.DPDK-ADVANCED.md index 577e412..9ce507b 100644 --- a/INSTALL.DPDK-ADVANCED.md +++ b/INSTALL.DPDK-ADVANCED.md @@ -16,6 +16,7 @@ OVS DPDK ADVANCED INSTALL GUIDE 11. [Jumbo Frames](#jumbo) 12. [Vsperf](#vsperf) 13. [Port Hotplug](#hotplug) +14. [Vdev Support](#vdev) ## <a name="overview"></a> 1. Overview @@ -898,6 +899,28 @@ This feature is not supported with VFIO and could not work with some NICs, please refer to the [DPDK Port Hotplug Framework] in order to get more information. +## <a name="vdev"></a> 14. Vdev Support (experimental) + +DPDK provides drivers for both physical and virtual devices. Physical 'dpdk' +devices are added to OVS by specifying a valid PCI address in 'dpdk-devargs'. +Virtual 'dpdkvdev' devices which do not have PCI addresses can be added using a +different format for 'dpdk-devargs'. + +Typically, the format expected is 'eth_<driver_name><x>' where 'x' is a +number between 0 and RTE_MAX_ETHPORTS -1 (31). + +For example to add a 'dpdkvdev' port that uses the 'null' DPDK PMD driver: + +`ovs-vsctl add-port br0 null0 -- set Interface null0 type=dpdkvdev options:dpdk-devargs=eth_null0` + +More information on the different types of virtual DPDK PMDs can be found in +the DPDK documentation: + +http://dpdk.org/doc/guides/nics/overview.html + +Not all DPDK virtual PMD drivers have been tested and verified to work. As such +support for these and the 'dpdkvdev' port type can be considered experimental. + Bug Reporting: -------------- diff --git a/NEWS b/NEWS index 4e3df06..fbd4a23 100644 --- a/NEWS +++ b/NEWS @@ -36,6 +36,7 @@ Post-v2.6.0 * Port Hotplug is now supported. * DPDK physical ports can now have arbitrary names. The PCI address of the device must be set using the 'dpdk-devargs' option. + * New 'dpdkvdev' port type for use with virtual DPDK Poll Mode Drivers. - Fedora packaging: * A package upgrade does not automatically restart OVS service. diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c index 816e0a6..d9cbc08 100644 --- a/lib/netdev-dpdk.c +++ b/lib/netdev-dpdk.c @@ -1062,7 +1062,7 @@ netdev_dpdk_get_config(const struct netdev *netdev, struct smap *args) } static void -netdev_dpdk_process_devargs(struct netdev_dpdk *dev, const char *devargs) +netdev_dpdk_process_pdevargs(struct netdev_dpdk *dev, const char *devargs) { int i = 0; struct rte_eth_dev_info info; @@ -1105,6 +1105,26 @@ netdev_dpdk_process_devargs(struct netdev_dpdk *dev, const char *devargs) } static void +netdev_dpdk_process_vdevargs(struct netdev_dpdk *dev, const char *devargs) +{ + uint8_t port_id; + + if (!rte_eth_dev_attach(devargs, &port_id)) { + /* Attach successful */ + VLOG_INFO("Port '%s' attached to DPDK as virtual device", + dev->up.name); + dev->port_id = port_id; + } else { + /* Attach unsuccessful */ + return; + } + + rte_eth_dev_stop(dev->port_id); + dev->socket_id = rte_eth_dev_socket_id(dev->port_id); + dpdk_eth_dev_init(dev); +} + +static void dpdk_set_rxq_config(struct netdev_dpdk *dev, const struct smap *args) OVS_REQUIRES(dev->mutex) { @@ -1171,7 +1191,28 @@ netdev_dpdk_set_config(struct netdev *netdev, const struct smap *args) if (!rte_eth_dev_is_valid_port(dev->port_id)) { new_devargs = smap_get(args, "dpdk-devargs"); if (new_devargs && strlen(new_devargs)) { - netdev_dpdk_process_devargs(dev, new_devargs); + netdev_dpdk_process_pdevargs(dev, new_devargs); + } + } + + ovs_mutex_unlock(&dev->mutex); + + return 0; +} + +static int +netdev_dpdk_vdev_set_config(struct netdev *netdev, const struct smap *args) +{ + struct netdev_dpdk *dev = netdev_dpdk_cast(netdev); + const char *new_devargs; + + ovs_mutex_lock(&dev->mutex); + dpdk_set_rxq_config(dev, args); + + if (!rte_eth_dev_is_valid_port(dev->port_id)) { + new_devargs = smap_get(args, "dpdk-devargs"); + if (new_devargs && strlen(new_devargs)) { + netdev_dpdk_process_vdevargs(dev, new_devargs); } } @@ -3222,6 +3263,22 @@ static const struct netdev_class dpdk_vhost_client_class = netdev_dpdk_vhost_client_reconfigure, netdev_dpdk_vhost_rxq_recv); +static const struct netdev_class dpdk_vdev_class = + NETDEV_DPDK_CLASS( + "dpdkvdev", + NULL, + netdev_dpdk_construct, + netdev_dpdk_destruct, + netdev_dpdk_vdev_set_config, + netdev_dpdk_set_tx_multiq, + netdev_dpdk_eth_send, + netdev_dpdk_get_carrier, + netdev_dpdk_get_stats, + netdev_dpdk_get_features, + netdev_dpdk_get_status, + netdev_dpdk_reconfigure, + netdev_dpdk_rxq_recv); + void netdev_dpdk_register(void) { @@ -3229,4 +3286,5 @@ netdev_dpdk_register(void) netdev_register_provider(&dpdk_ring_class); netdev_register_provider(&dpdk_vhost_class); netdev_register_provider(&dpdk_vhost_client_class); + netdev_register_provider(&dpdk_vdev_class); } diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml index 16318c5..39c69a9 100644 --- a/vswitchd/vswitch.xml +++ b/vswitchd/vswitch.xml @@ -2304,8 +2304,13 @@ <column name="options" key="dpdk-devargs" type='{"type": "string"}'> <p> - Specifies the PCI address of a physical dpdk device. - Only supported by 'dpdk' devices. + For 'dpdk' ports specifies the PCI address associated with the port. + For 'dpdkvdev' ports specifies the the virtual driver to be used for + the particular port and typically takes the form of + eth_<driver_name><x> where 'driver_name' is a valid + virtual DPDK PMD driver name and where 'x' lies in the range of 0 to + RTE_MAX_ETHPORTS-1. + Not supported by port types other than those mentioned above. </p> </column> -- 2.4.3 _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
